As the dimensions within integrated circuits continue to shrink, the resolution limit of the radiation used to expose photoresist continues to be the final barrier to shrinking dimensions even further. The introduction of the phase shift mask has allowed this barrier to be pushed towards smaller and smaller features. The principle on which a phase shift mask operates is to bring about the destructive interference of the diffraction fringes normally found at the edge of an image, thereby improving image contrast.
For lithography at 365 nm (i-line) and 248 nm (for example from a KrF laser) the attenuated phase shifting mask (APSM) has been widely used because of its ease of design and fabrication. A wide range of materials suitable for use as APSMs are available, including silicon oxide, chromium oxide, silicon nitride, silicon carbo nitride, and molybdenum silicon oxynitride. Application is usually focussed on contact layers due to alleviation of defect problems in dark field masks. It can push the resolution of contact printing one generation ahead since aerial image contrast is enhanced.
When, however, the wavelength of the resist exposure radiation is reduced still further to 193 nm (ArF laser), the materials listed above are no longer adequate and at least one replacement must be found if efficient phase shift masks for use at 193 nm are to be developed. This is because the refractive index (n) and the extinction coefficient are both functions of wavelength.
The two key optical properties that must be possessed by a material if it is to be suitable for use as a phase shift mask at 193 nm are that its transmittance (at a thickness corresponding to a 180 degree phase shift) must be between about 4 and 15% while its reflectance must be about 15%. These properties, in turn, depend on the refractive index and extinction coefficient of the material in question (at 193 nm).
Referring now to FIG. 1, we show three curves, measured at 193 nm, of extinction coefficient (attenuation of transmitted radiation due to both absorption and scattering) as a function of refractive index. Curve 11 is for a transmittance of 4% (corresponding to a relatively thick film), curve 12 is for a transmittance of 15% (corresponding to a relatively thin film), and curve 13 is for a reflectance of 15% (independent of thickness).
Thus for a material to be suitable for use as a phase shift mask at 193 nm its optical properties must be such that it is located within the area marked as 15 in FIG. 1, as close to line 13 as possible. Thus, one set of optimal properties would be an extinction coefficient of about 0.4 and a refractive index of about 2.5.
The point marked as SiN.sub.x in FIG. 1 indicates where non-stoichiometric silicon nitride (having the desired properties) falls on this plot. It is clear from this that non-stoichiometric silicon nitride is what is needed for a 193 nm APSM. This conclusion is confirmed by the data shown in FIG. 2 which is for stoichiometric silicon nitride. As can be seen, at 193 nm, the refractive index is 2.65 (curve 21) while the extinction coefficient is about 0.17.
We refer now to FIG. 3 which is similar to FIG. 2 in that curves of refractive index and extinction coefficient, as a function of wavelength, for silicon nitride are also plotted. These samples were deposited by means of sputtering and have been reported by B. W. Smith et al. in SPIE 1997, vol. 3051, pp. 236-244. In the course of transferring silicon nitride from a target to a substrate, some of the nitrogen was lost so the films ended up as silicon rich. The effects of this departure from stoichiometry are reflected in the values for refractive index and extinction coefficient at 193 nm. These were about 2.37 and 0.38, respectively, much closer to the ideal values of 2.5 and 0.4 than could be achieved with non-stoichiometric films.
While these optical values for silicon rich films are attractive, other properties of the silicon rich films, such as UV stability and etch characteristics, make them unattractive for use within an integrated circuit manufacturing process. Other attempts to modify the optical properties of silicon nitride by varying the silicon/nitrogen ratio have also been reported but it has turned out that silicon rich silicon nitride is much easier to form than a nitrogen rich version. For example, Hess et al. (U.S. Pat. No. 4,863,755 September 1989) show how PECVD may be used to generate silicon nitride films. They use an organic silicon/nitrogen source in conjunction with hydrogen or ammonia. Some control of stoichiometry is achieved by varying process conditions but the films also include significant quantities of carbon and oxygen.
Hubler et al. (U.S. Pat. No. 5,015,353 May 1991) teach how films of composition Si.sub.1-x N.sub.x, where x is controlled to be between 0 and 0.57, can be formed. Thus any departure from stoichiometry that they achieved was always on the side of being silicon rich.
Dowben (U.S. Pat. No. 5,468,978 November 1995) also teaches that the stoichiometry of films produced by PECVD may be controlled by varying concentration of precursor materials and other process parameters. However the material produced was boron carbide so a direct comparison with silicon nitride is not possible.